These are visualizations for the observations, predictions (output of trained model), and the residuals. the x axis is the log of the time mark and y is the detector angle in degrees. The first four rows are for monodisperse samples, and the last row is the mix 80% 200nm 20% 1000nm. the trained models are for a single source (k=1) and the parameters are written in the annotation on the left side of the row. It looks like the the model is working well for the monodisperse samples when we inspect that the mu values are in proportion to the radius reciprocal: 100nm: 0.3797 200nm: 0.2153 500nm: 0.0739 1000nm: 0.0368 the last mixture shows that the single-source model finds the 200nm mu, since it comprises 80% (volume?) of the sample.